Images from a single API call.

Two models behind one JSON endpoint. Text-to-image, reference edits, masks and transparent output. Pay only for what you generate.

Read the docs OpenAPI spec

POST https://zuki.pw/v1

[{
  "taskType": "imageInference",
  "model": "zuki-image-1.0",
  "positivePrompt": "Concrete tower, low angle, B&W",
  "width": 1280, "height": 1024
}]

200

{ "imageURL": "…/v1/images/84a325ac.jpg" }

Made through the API

A few results from both models, each from a single request.

Two models, one set of parameters

Start fast with the base model and switch to Max for detailed, iterative work. Changing the model is a one-word edit.

zuki-image-1.0

Zuki Image 1.0

Fast generation and editing with natural lighting and fine texture. Built for high-volume product, creator and prototype work.

Typical time: ~20 s
Max side: 3840 px
Formats: JPG, PNG, WEBP

zuki-image-1.0-max

Zuki Image 1.0 Max

Premium generation and editing with precise control. Tighter control over detailed and iterative edits, for campaign creative and product imagery.

Typical time: 30-70 s
Max side: 3840 px
Formats: JPG, PNG, WEBP

What the API covers

Text to image

Prompts up to 32,000 characters. Any size from about 0.65 to 8.3 megapixels, up to 3840 px on a side.

Reference images

Up to 16 references per task to keep a subject, product or style consistent.

Masked editing

Paint the area to change in white. Everything else in the image stays untouched.

Transparent output

PNG or WEBP with alpha for stickers, cut-outs and product shots.

Async and webhooks

Queue a task and poll it, or receive every image on your own endpoint.

Pay per image

An estimate is reserved while a task runs. Failed generations cost nothing.

First image in under a minute

Your key starts with zk-. Paste it into the quickstart and send the request.

Read the docs